A system, comprising a computer that includes a processor and a memory, the memory storing instructions executable by the processor to input a red-green-blue (RGB) image and an eccentricity image to a neural network which outputs a located object based on combining the RGB image and the eccentricity image, wherein the eccentricity image is based on a per-pixel rolling average and a per-pixel rolling variance over a moving window of k video frames. The memory can further include instructions executable by the processor to receive the located object at a computing device included in one or more of a vehicle or a traffic information system.
Legal claims defining the scope of protection, as filed with the USPTO.
2. The system of claim 1, wherein the RGB image is fused with the eccentricity image by addition, multiplication or concatenation.
3. The system of claim 1, wherein the neural network is one or more of a you only look once (YOLO) neural network, a faster-region-based convolutional neural network (Faster-RCNN), and a single shot detector (SSD).
4. The system of claim 1, wherein the instructions further include instructions to re-size the eccentricity image before combining with the RGB image.
5. The system of claim 1, wherein the instructions further include instructions to combine the eccentricity image with the RGB image by one or more of per-pixel addition, per-pixel multiplication and per-pixel concatenation.
6. The system of claim 5, wherein the instructions further include instructions to process the eccentricity image with a second neural network prior to concatenation with the RGB image.
7. The system of claim 1, wherein the instructions further include instructions to combine the RGB image and the eccentricity image before processing with the neural network.
8. The system of claim 1, wherein the neural network includes a neural network body and a neural network head, and the instructions further include instructions to combine the eccentricity image with RGB features following feature detection in the RGB image by the neural network body and prior to processing by the neural network head.
9. The system of claim 1, wherein the instructions further include instructions to receive, at a vehicle, the located object and operating the vehicle based on the located object.
10. The system of claim 9, wherein the located object is measured in global coordinates based on intersecting a video camera field of view and a two-dimensional plane parallel to a roadway.
12. The method of claim 11, wherein the RGB image is converted to grayscale before combining with the eccentricity image.
13. The method of claim 11, wherein the neural network is one or more of a you only look once (YOLO) neural network, a faster-region-based convolutional neural network (Faster-RCNN), and a single shot detector (SSD).
14. The method of claim 11, wherein the eccentricity image is re-sized before the combining with the RGB image.
15. The method of claim 11, wherein the eccentricity image is combined with the RGB image by one or more of per-pixel addition, per-pixel multiplication and per-pixel concatenation.
16. The method of claim 15, wherein the eccentricity image processed with a second neural network prior to concatenation with the RGB image.
17. The method of claim 11, further comprising combining the RGB image and the eccentricity image before processing with the neural network.
18. The method of claim 11, wherein the neural network includes a neural network body and a neural network head, and the eccentricity image is combined with RGB features following feature detection in the RGB image by the neural network body and prior to processing by the neural network head.
19. The method of claim 11, further comprising receiving, at a vehicle, the located object and operating the vehicle based on the located object.
20. The method of claim 19, wherein the located object is measured in global coordinates based on intersecting a video camera field of view and a two-dimensional plane parallel to a roadway.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
May 24, 2019
October 4, 2022
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.